0
$\begingroup$

Apologize for insufficient info in my previous version of this question. So, I am comparing patient survival in a retrospective study between two different groups, using SPSS. Now the difference came significant by KM, so I tried to adjust multiple covariates for confounding effect in a Cox model. I did a backward Wald test by putting main group in block 1 and co-variates in block 2. The output window gave me this:

Case Processing Summary         
        N   Percent
Cases available in analysis Event   6065    10.3%
    Censored    24860   42.1%
    Total   30925   52.3%
Cases dropped   Cases with missing values   28189   47.7%
    Cases with negative time    0   0.0%
    Censored cases before the earliest event in a stratum   0   0.0%
    Total   28189   47.7%
Total       59114   100.0%
Dependent Variable: patient survival time (days)            


Omnibus Tests of Model Coefficients
-2 Log Likelihood
115511.442 Omnibus Tests of Model Coefficientsi                                     
Step  -2 LL  Overall (score)   Change From Prev Step   Change From Prev Block       
        Chi-square  df  Sig.    Chi-square  df  Sig.    Chi-square  df  Sig.
1a  113504.047  1998.887    23  .000    2006.433    22  .000    2006.433    22  .000
2b  113504.119  1998.833    21  .000    .072    2   .965    2006.361    20  .000
3c  113504.185  1998.423    19  .000    .066    2   .968    2006.295    18  .000
4d  113504.302  1998.059    17  .000    .118    2   .943    2006.177    16  .000
5e  113506.039  1996.799    15  .000    1.737   2   .420    2004.441    14  .000
6f  113507.146  1986.667    14  .000    1.108   1   .293    2003.333    13  .000
7g  113516.047  1978.462    7   .000    8.900   7   .260    1994.433    6   .000
8h  113518.520  1972.010    6   .000    2.473   1   .116    1991.960    5   .000
aStep 8 TX_MONTHJ   0.011   0.026   0.183   1   0.669   1.011   0.961   1.063
    AGE 0.043   0.001   1483.638    1   0   1.044   1.042   1.047
    AGE_GROUP   0.701   0.119   34.677  1   0   2.015   1.596   2.544
    AGE_DON 0.006   0.001   56.755  1   0   1.006   1.004   1.007
    NDIAL2_TRR(1)   0.129   0.041   9.873   1   0.002   1.138   1.05    1.233

I interpret this as saying that no difference was observed, but I'm confused about what to report. Could you also verify if this test looks good with the likelihood ratio so big?

$\endgroup$
0
1
$\begingroup$

If you are writing a paper, the journal probably has a style guide. Often, these show what to report. (This is assuming your question is mostly about which statistics to report and how to report them, not how to do the analysis).

$\endgroup$
0
0
$\begingroup$

Of course this depends on your objectives. First of all you might want to show that you checked the validity of the proportional hazards assumption. But usually the model is used to compare two or more groups. Then you compute confidence intervals for the hazard ratio. If the confidence interval for a hazard ratio does not contain 1 then you can conclude that one group has better survival than the other. This is commonly what you want to show when comparing a treatment to a control when the treatment is expected to improve survival.

If you have a different objective there could be a different answer. That is why moderator chl suggested that your question is too vague.

$\endgroup$
1
  • $\begingroup$ Apologize for insufficient info. So, I am comparing patient survival in a retrospective study between two different groups, using SPSS. Now the difference came significant by KM, so i tried to adjust multiple co-variates for confounding effect in Cox-Model.I did backward Wald by putting main group in block 1 and co-variates in block 2. The output window gave me $\endgroup$
    – user13591
    Aug 25 '12 at 22:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.